CKM targets the activator Gcn4 for phosphorylation

Several transcription activators, particularly ones involved in response to stress and nutrient deficiency, are known targets of the CKM kinase Cdk8 [77], committing them to proteasomal degradation. At upstream activating sequences, chromatin-bound activators recruit the Mediator complex, which in turn contacts the PIC on promoters to activate transcription by stimulating promoter escape. In in vitro biochemical binding terms, this means that activators at upstream activating sequences are part of a ternary interaction; contacting DNA on one side, and Mediator on the other. We asked whether phosphorylation of activators by CKM has an effect on this DNA-activator-Mediator ternary interaction, schematically depicted in figure 5.16 A.

Apart from global turnover effects, we asked what are the direct effects of CKM-dependent phosphorylation of activators. To address that, we purified the yeast activator General control protein 4 (Gcn4), one of the best-characterized yeast activators. Gcn4 is a typical activator, with a distinct structured DNA-binding domain, and a long, disordered activation domain. We set up a kinase assay to test in vitro phosphorylation of Gcn4 by CKM. Figure 5.16 B shows that Gcn4 phosphorylation by

Cdk8 can be recapitulated in vitro, and can occur on free or DNA-bound Gcn4 as a substrate, with no visible effect on the kinetics of the phosphorylation reaction.

Having established CKM-dependent phosphorylation of Gcn4 in vitro, we went on to investigate the effect this may have on the DNA-activator-Mediator ternary interaction. To approach this problem, we divided it into its two binary components.

CKM phosphorylation of Gcn4 impairs its genomic binding

First, we investigated the effect of CKM phosphorylation on the ability of activators to bind their cognate binding sites on DNA (figure 5.17). To do that, we employed a combination of complementary assays. On an immobilized template assay, we used a DNA template modeled on the upstream sequence adjacent to the yeast HIS4 promoter, which contains five binding sites for the activator Gcn4 in the natural sequence. We used a truncated version, which contained only three Gcn4 binding sites with a biotinylation on the 3’ side. The DNA scaffold was immobilized on streptavidin beads, and Gcn4 was bound to the DNA scaffold, the unbound excess Gcn4 was washed away, and then the Gcn4-DNA complex was incubated with CKM(A or KD), and ATP (figure 5.17 B), and then eluted by restriction digestion.

Strikingly, we saw a remarkable reduction in affinity upon phosphorylation, indicating that CKM phosphorylation can actively dislodge Gcn4 from its genomic binding site.

This could be confirmed by measurement of the affinity of CKM-phosphorylated versus unphosphorylated Gcn4 to a fluorescein-labelled DNA scaffold containing a Gcn4 binding site by fluorescence anisotropy, which showed that unphosphorylated Gcn4 had close to double the affinity to DNA than its phosphorylated counterpart (figure 5.17 C). Please note that the Kd values presented here are not meaningful in absolute terms, but only in relative terms, due to the nature of the experimental setup, which allowed only relative concentrations to be determinable (please refer to the methods section for experimental details).

Mapping the phosphorylation sites by phosphopeptide enrichment and mass spectrometry revealed sites within the disordered activation domain of Gcn4, as well as one site in the DNA-binding domain (figure 5.17 E). Mapping this site onto the crystal structure of the Gcn4 DNA-binding domain (DBD) [108], showed that it occurs within the highly hydrophobic interface of the leucine zipper formed by dimerization of the DBDs of two Gcn4 molecules, providing a potential explanation for the loss of affinity to DNA upon phosphorylation.

Gcn4 and other activators with low complexity activation domains have been shown to undergo liquid-liquid phase separation in vitro and in vivo [52]. To test the effect of DNA-binding on droplet formation, we mixed Gcn4 with or without fluorescein-labelled DNA containing a Gcn4 binding site with dextran, as previously described.

DNA was incorporated into Gcn4 droplets, as evident by detection of green fluorescence within the droplets, and resulted in markedly bigger droplets than Gcn4 alone (figure 5.17 D). Having separately shown that CKM phosphorylation of Gcn4 impairs its DNA binding, and that DNA binding enhances liquid droplet formation, it can be inferred that CKM phosphorylation attenuates Gcn4-DNA phase separation.

CKM phosphorylation impairs the interaction of Mediator with activators and dissolves activator/mediator liquid droplets

Next, we looked at the interaction between activators and Mediator (figure 5.18). Our recombinant cMed does not contain the tail module, which is responsible for the majority of activator-Mediator interactions in yeast, and contains many characteristic low complexity sequences, which impart a high degree of structural plasticity, which is conserved from yeast to human. Technically, this also means that this Mediator

module is difficult to recombinantly express and purify. Owing to the difficulty of performing biochemical studies with the minute amounts of purified endogenous Mediator that contains a tail module, we resorted to recombinantly expressing and purifying only the part of the complex that has been reported to interact with the Gcn4 activator. We capitalized on the wealth of knowledge available about the Gcn4-Mediator interaction, which has been meticulously characterized by heteronuclear single quantum coherence spectroscopy and XL-MS [109]. This study showed that Gcn4 binds the Mediator tail subunit Med15 at four distinct structured binding domains called KIX, ABD1, ABD2 and ABD3, which are interspersed between disordered polyglutamine (polyQ) and polyglutamine/polyalanine (polyQ/A) linkers.

The binding of the Gcn4 activation domain to all of the Med15 structured domains led to the description of this interaction as a “fuzzy free-for-all” rather than a defined interaction.

Knowing that the four Med15 structured domains are sufficient for Gcn4 binding, we purified a Med15 variant, Med15_KIX123, where we removed all the polyQ and polyQ/A linkers and concatenated the four binding domains together. To test the effect of CKM-phosphophorylation of Gcn4 on the binding of Gcn4 to Med15, we performed an electrophoretic mobility shift assay (EMSA), in which a constant amount of phosphorylated or unphosphorylated GFP-tagged Gcn4 was titrated against increasing concentrations of Med15_KIX123. We found that phosphorylation of Gcn4 weakens this interaction (figure 5.18 B).

It has also been reported that Gcn4 and Med15 undergo liquid-liquid phase separation and colocalize in the same droplets [52]. To test the effect of Gcn4 phosphorylation on Med15-Gcn4 phase separation (figure 5.18 C), we reproduced the experiment in this study, by purifying the N-terminal half of Med15 (residues 6-651), this time with its polyQ and polyQ/A linkers, with an mCherry fusion tag. We purified mCherry-Med15 as in [52]. This purification is crude - expression levels are minute due to the extensive polyQ and Q/A linkers - but the presence of a fluorescent tag allowed unambiguous detection of the protein.

We incubated a 1:1 mix of GFP-Gcn4 and mCherry-Med15 with CKM(KD or A) and ATP, and then added dextran and visualized the formed phase-separared droplets under a fluorescence microscope. Our results show that phosphorylation by CKM attenuated these droplets (figure 5.18 C).

Taken together, the weakening of the Med15-Gcn4 interaction, and the shrinkage of Med15-Gcn4 droplets, upon CKM phosphorylation, is an indication of CKM-mediated release of Mediator at upstream activating sequences, and a resolution of local Mediator-activator foci (figure 5.18 D).